Readability theme generation

Patent metadata

Field	Value
Publication number	US-12481818-B2
Application number	US-202318168187-A
Country	US
Kind code	B2
Filing date	Feb 13, 2023
Priority date	Feb 13, 2023
Publication date	Nov 25, 2025
Grant date	Nov 25, 2025

Abstract

Techniques are disclosed for readability theme generation. The techniques include obtaining reading formats and generating reading format digital images based on the obtained reading formats. The reading format digital images are encoded using a trained machine learning model as perceptual embeddings. The perceptual embeddings are clustered into reading format clusters and readability themes are determined based on the reading format clusters.

First claim

We claim:

1 . A method comprising:
obtaining a plurality of reading formats; generating a plurality of reading format digital images based on the plurality of reading formats; inputting the plurality of reading format digital images to a trained machine learning model, wherein the trained machine learning model includes a trained convolutional neural network-based machine learning model; encoding, using the trained convolutional neural network-based machine learning model, the plurality of reading format digital images as a plurality of perceptual embeddings; clustering the plurality of perceptual embeddings into a plurality of reading format clusters by:
determining an optimal number of clusters based on one or more clustering quality metrics that measure a similarity of perceptual embeddings in each cluster, and
clustering the plurality of perceptual embeddings into the optimal number of clusters; and
determining a set of readability themes based on the plurality of reading format clusters, wherein each readability theme is based on a representative perceptual embedding of a corresponding reading format cluster.

2 . The method of claim 1 , further comprising:
providing graphical user interface controls in a graphical user interface, the graphical user interface controls for selecting a readability theme from among the set of readability themes to apply to text of a document.

3 . The method of claim 1 , wherein obtaining the plurality of reading formats is based on:
receiving the plurality of reading formats from a plurality of reading devices used by a cohort of users.

4 . The method of claim 1 , wherein generating the plurality reading format digital images based on the plurality of reading formats is based on:
outputting text for display in a virtual frame buffer; capturing, from one or more virtual frame buffers, a set of digital image screenshots of text formatted in accordance with the plurality of reading formats; and extracting the plurality of reading format digital images from the set of digital image screenshots.

5 . The method of claim 1 , wherein the trained machine learning model comprises a trained convolutional neural network-based machine learning model, and wherein encoding, using the trained machine learning model, the plurality of reading format digital images as the plurality of perceptual embeddings is based on:
inputting the plurality of reading format digital images to the trained convolutional neural network-based machine learning model; and obtaining the plurality of perceptual embeddings from the trained convolutional neural network-based machine learning model.

6 . The method of claim 1 , wherein the trained machine learning model comprises a trained convolutional neural network-based machine learning model; and wherein encoding, using the trained machine learning model, the plurality of reading format digital images as the plurality of perceptual embeddings is based on:
training the convolutional neural network-based machine learning model to generate the plurality of perceptual embeddings based on the plurality of reading format digital images.

7 . The method of claim 1 , wherein clustering the plurality of perceptual embeddings into the plurality of reading format clusters is based on:
determining an optimal number of K clusters based on the plurality of perceptual embeddings; and clustering the plurality of perceptual embeddings into K clusters.

8 . The method of claim 1 , wherein determining the set of readability themes based on the plurality of reading format clusters is based on:
determining a respective centroid of each reading format cluster of the plurality of reading format clusters; selecting a respective perceptual embedding from each reading format cluster of the plurality of reading format clusters based on the respective centroid determined for the reading format cluster; and determining a respective readability theme for each reading format cluster of the plurality of reading format clusters based on the respective perceptual embedding selected from the reading format cluster.

9 . The method of claim 1 , wherein a reading format image of the plurality of reading format images has a size determined based on a predetermined viewing distance and a predetermined visual angle.

10 . The method of claim 1 , wherein a reading format of the plurality of reading formats comprises a font identifier, a line spacing setting, a character spacing setting, and a word spacing setting.

11 . The method of claim 1 , wherein:
each reading format digital image of the plurality of reading format digital images depicts text formatted in accordance with a respective reading format of the plurality of readings formats; and wherein text depicted in the plurality of reading format digital images has a same font size across the plurality of reading format digital images and is normalized for x-height across the plurality of reading format digital images.

12 . A system comprising:
one or more memory components; and one or more processing devices coupled to the one or more memory components, the one or more processing devices to perform operations comprising:
obtaining a plurality of reading formats;
generating a plurality of reading format digital images based on the plurality of reading formats;
inputting the plurality of reading format digital images to a trained machine learning model, wherein the trained machine learning model includes a trained convolutional neural network-based machine learning model;
encoding, using the trained convolutional neural network-based machine learning model, the plurality of reading format digital images as a plurality of perceptual embeddings;
clustering the plurality of perceptual embeddings into a plurality of reading format clusters by:
determining an optimal number of clusters based on one or more clustering quality metrics that measure a similarity of perceptual embeddings in each cluster, and
clustering the plurality of perceptual embeddings into the optimal number of clusters; and

determining a set of readability themes based on the plurality of reading format clusters, wherein each readability theme is based on a representative perceptual embedding of a corresponding reading format cluster.

13 . The system of claim 12 , the one or more processing devices to further perform operations comprising:
providing graphical user interface controls in a graphical user interface, the graphical user interface controls for selecting a readability theme from among the set of readability themes to apply to text of a document.

14 . The system of claim 12 , wherein obtaining the plurality of reading formats is based on:
receiving the plurality of reading formats from a plurality of reading devices used by a cohort of users.

15 . A non-transitory computer-readable medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising:
receiving a plurality of readability themes, wherein the plurality of readability themes is generated based on:
inputting a plurality of reading format digital images to a trained machine learning model, wherein the trained machine learning model includes a trained convolutional neural network-based machine learning model;
encoding, using the trained convolutional neural network-based machine learning model, the plurality of reading format digital images as a plurality of perceptual embeddings;
clustering the plurality of perceptual embeddings into a plurality of reading format cluste

Assignees

• Adobe Inc

Inventors

• Cai Tianyuan
• Bylinskii Zoya
• Kraley Michael
• Niklaus Aleena

Classifications

• G06F3/0482

  Interaction with lists of selectable items, e.g. menus · CPC title

• G06V30/19107

  Clustering techniques · CPC title

• G06V10/82 Primary

  using neural networks · CPC title

• G06V30/19147

  Obtaining sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06F3/04847

  Interaction techniques to control parameter settings, e.g. interaction with sliders or dials · CPC title